.Impact van pensioenhervormingen op de participatiegraad van oudere werknemers in de EU

8/14/2019 .Impact van pensioenhervormingen op de participatiegraad van oudere werknemers in de EU

1/44


2/44

Economic Papers are written by the Staff of the Directorate-General for Economic and Financial

Affairs, or by experts working in association with them. The Papers are intended to increase awareness

of the technical work being done by staff and to seek comments and suggestions for further analysis.

The views expressed are the authors alone and do not necessarily correspond to those of the European

Commission. Comments and enquiries should be addressed to:

European Commission

Directorate-General for Economic and Financial Affairs

Publications

B-1049 Brussels

Belgium

E-mail: [email protected]

This paper exists in English only and can be downloaded from the website

http://ec.europa.eu/economy_finance/publications

A great deal of additional information is available on the Internet. It can be accessed through theEuropa server (http://europa.eu )

KC-AI-09-385-EN-N

ISSN 1725-3187

ISBN 978-92-79-13360-2

DOI 10.2765/28805

European Communities, 2009
http://ec.europa.eu/economy_finance/publicationshttp://europa.eu/http://europa.eu/http://ec.europa.eu/economy_finance/publications


3/44

Assessing the short-term impact of pension reforms on older workers' participation

rates in the EU: a diff-in-diff approach

Alfonso Arpaia** - Kamil Dybczak *- Fabiana Pierini*

1. Introduction

The performance of the European labour markets improved significantly during the second half

of the 1990s (AER 2003). After having reached a peak in 1994, the unemployment rate started

gradually to decline while both the employment and the participation rates kept rising. With

increases of more than 8 and 7 percentage points, respectively for the employment and the

participation rates, the female and the older workers were the most dynamic components. These

improvements reflect long-term changes in the socio-economic behaviour such as a different

attitude toward female employment and participation, improved health and working conditions

which induce to retire at older ages. Yet, they took place in response to the reforms implemented

during the period (e.g. ECB, 2007). The last decade witnessed important changes in European

pension systems. Up to 1995, only few countries implemented pension reforms. By 2006, almost

every European country had enacted reforms of the pension system. This richness of reforms

across countries and time of their occurrence can be used to conduct a "policy experiment" of

the effects of pension reforms on the participation rates of people aged between 50 and 64 years.

Each policy intervention is considered as a discrete event that occurred at a specific time for each

country. The event-study compares the value of one variable of interest after a certain reform orlegislation has taken place with its value before such change has occurred. To control for other

determinants not related to specific policy interventions, the findings of before-after comparison

are compared with a control group made of those countries which did not implement a reform at

least in one year covered by the sample period. With the event-study approach we will verify

** DG ECFIN, European Commission and IZA . * DG ECFIN, European Commission.An early version of this paper was presented at the 11 th Banca d'Italia workshop on Public FinancePension Reform,Fiscal Policy and Economic Performance, Perugia S.A.DI.B.A.26-28 March 2009. We

have benefited from comments by conference participants, we would like to thank Lukas Reiss, in particular. The analysis and conclusions set forth are those of the authors and do not indicateconcurrence by other members of the European Commission.

3


4/44

whether after pension reforms the participation rate rises.1 Thus, we analyse the impact of

pension reforms on participation rates of different age/sex groups of elderly workers by

contrasting changes in participation rates in reforming vs non-refroming countries.

The paper is organised as follows. Section 2 presents the main stylised facts. Section 3 briefly

reviews the main theoretical explanations of the observed trends in participation, while section 4

discusses the effects of pension reforms on the average retirement age. Section 5 gives an

overview of the reforms undertaken in the EU between 1997 and 2007. Section 6 presents the

empirical finding of the effects of recently introduced pension reforms on the older workers'

activity rates. Section 7 discusses the policy implications and possible follow up.

2. Stylised facts: main developments in older workers' participation rates

Life expectancy has significantly increased in developed countries, mainly thanks to improved

living standards, working conditions and health care. In the early 1980s the average life

expectancy stood at around 75 years to reach 80 in 2006; for few new Member States it hovered

around the EU average of 26 years earlier. (Table 4).

Work has become less physically demanding, population much healthier and long-lived. Even so,

as documented, among others, by Palmer (1999), Samwick (2002), and Boeri et al (2001), there

has been a significant decline in the participation rate of elderly people, which reversed its

negative trend only in recent years. The dramatic difference in the time pattern across men and

women (Graph 1 ) often gets unnoticed. For several countries, the activity rate of men aged

between 55 and 64 appears often U shaped, with decline in participation at least until the mid

1990s. For the 50-54 age group, rates appear more stable and the decline relatively more limited;

there are significant exceptions to this pattern such as. the participation rates of Belgian and

Italian men aged 50-54, rapidly converging to the highest rates. Despite country specific labour

force histories, the broad trend of a shrinking labour supply of male aged 50+ remains. Thus,even though men live longer than before, they leave the labour market earlier.

Conversely, women, especially those aged less then 60, have a steadily rising participation, and it

is not rare to find countries where female rates almost doubled in 10 years only. The change over

time in the age profile of the participation rates confirms that the major modifications in the

participation behaviour occurred in the case of women, at age below 59, and especially in their

early 50s. Without these modifications, several countries would have had in 2007 activity rates

1 The event-study method has been applied to study market response to changes in the law, both as a result of court

4


5/44

hovering around the level of twenty years earlier. As a consequence of these differentiated

patterns in the participation rates by sex, the average age at which people retire has changed only

to a minor extent (Table 5).

Graph 2 displays the age profile of the exit rate from the labour market for selected countries for

the mid 80s, the early 1990s and the first half of 2000s. This rate is calculated as the conditional

probability of an age cohort of not staying in the labour market at age h.2 Spikes can be observed

at about the statutory retirement age for all countries and, for some, at the age of early retirement.

There is also a clear difference in the exit rates by sex which reflects different statutory retirement

ages of men and women. Finally, there are recently significant changes in the age profile of the

exit rates in the recent years. The probability of leaving the labour market at ages just below 60

falls for both sexes in several countries. Even so, at the age of 60 there is a significant increase inthe probability of withdrawing from the labour market. Early exit from the labour market

remains high in Belgium, Germany, France, Italy and the Netherlands.

The patterns briefly described are the outcome of complex individual participation decisions

which are influenced by a variety of factors, includingsocial factors, such as longer schooling or

change in the role of women in households; demographic factors, including the decline of fertility

rates and modifications of the age structure; institutional factors, such as changes in the financial

incentives to retire early, in the eligibility conditions or in the availability of alternative early

retirement paths, (e.g. temporary access to disability and unemployment benefits before being

granted retirement benefits, Van Ours, 2006 for the Netherlands). Early- or pre-retirement

programmes were commonly used in the 70s and 80s to deal with industrial restructuring

(Brugiavini, 2001), high unemployment of older workers, low employment of young workers, or

as a labour cost saving strategies. Economic factors, such as the level of the unemployment rate,

the average income by household, the share of part-time employment in total employment or the

share of the services sector in the economy have also been invoked to explain the differences in

the participation rates across countries and over time.

3. What explains the main trends

Many economists have tried to solve the puzzle of higher life expectancy, less physically

demanding work and lower retirement ages. Two major factors have caused declining

participation rates of older workers (Diamond, 2005).

decisions and legislative reforms.2 In symbols if PR(h,t) is the participation rate at time t of cohort h, the exit rate is defined as 1-PR(h,t)/PR(h-1,t-1).

5


6/44

First, due to positive trend in real earnings, both the fraction of lifetime spent working dropped.

As the income effect from higher real earnings prevails on the substitution effect, higher real

incomes allows more hours for leisure, higher consumption and savings despite falling working

time. Thus, the increase in real wages has been the main determinant of the long-term decline in

the retirement age in industrialised countries (Bloom, Canning and Moore, 2007).

The increase in the lifespan has also produced a wealth effect because of the influence of

compound interest and wage growth, which reduce the proportion of life devoted to work.

Second, the rules establishing access to pension, public health and long-term care may have

influenced the individual decision to retire. As life expectancy increases it would be optimal to

postpone retirement age. However, the existence of social security programs translate into higher

savings and earlier labour market exits (e.g. Bloom, Canning, Mansfield and Moore, 2006 for alife-cycle model of the labour supply with endogenous retirement age and the social security

arrangement). Similarly, in a model with stochastic ageing among three age classes and

accumulation of human capital with two skill levels, Ljiungqvist (2007) shows that the non-

employment effect of taxation do not differ in complete and incomplete markets, with the tax

and benefit system affecting non-employment of low and high skilled respectively in complete

and incomplete markets.3 Using a panel for 12 countries, Gruber and Wise (2002) demonstrate

several disincentives for continued work for the elderly built in national social security schemes.

Many have noticed high exit rates at the first age at which one can retire and at the statutory

retirement age (e.g. Coile and Gruber, 2000 or Samwick, 1998). More generally, individuals able

to set aside enough funds are those that firstly retire, especially when they are allowed to use

benefits to "top-off" their retirement wealth.

Early retirement schemes can be characterised by several adverse mainly long-term effects

(Conde-Ruiz and Galasso, 2004). They can influence negatively the accumulation of human

capital of less-skilled workers, lower economic growth, and increase the dependency ratio and the

risks of financial imbalances when population ages. Using an overlapping generation model with

heterogeneous agents extended by voting, Conde-Ruiz and Galasso demonstrate why alternative

policies had not been realised even though they would have had less distortive impact upon the

economy. Their analysis provides a political economy explanation of the early retirement

schemes.

3 With incomplete markets fiscal policy impacts employment outcomes via the optimal allocation of individualwealth. As tax rates increase, skilled people can put aside enough funds to finance early retirement. At successivelyhigh rates, the low skilled will start to save up for early retirement.

6


7/44

4. Pension reforms and average retirement age

If expected income falls or life expectancy increases unexpectedly, a worker realises that his/her

planning horizon is extended and previous plans concerning the rest of his/her life should be

reassessed. Economic theory proposes three ways how to set up a new optimal plan. First, a

worker could reduce consumption during pre-retirement age and accumulate savings for later

stages of life. Second, a worker could reduce consumption spending during retirement age and

deplete lifetime savings more slowly. Third, a worker could decide to work longer to reach the

initial level of consumption. In addition, when there is only one earner in the family, the fall in his

or her expected income during retirement may induce the second earner to enter into the labour

market to keep unchanged the family consumption.4 The final impact on the participation rate

depends on how these effects influence the retirement decision.

Within a life-cycle framework, the retirement decision is a function of the lifetime streams of

earning, pensions and other sources of income (Mitchell and Fields 1981). Obviously rational

agents chose their optimal consumption pattern jointly with the amount of work they wish to

supply during their lifetime and the time at which they wish stop working. In a standard

competitive model with social security, taxes and benefits have distortionary effects on individual

consumption, savings and optimal retirement age (e.g. Seshinski, 1977). Thus, compared to an

economy with no benefits, social security benefits imply in equilibrium lower consumption and

lower retirement age. In the option value model (Stock and Wise, 1990), the work/retirement

decision is associated to the option ofcontinued work keeping the option to retire at a later stage. If the

expected value of working is worth more than the expected value of retiring, the individual

continues to work. If there are no expected gains from continued work, he would retire. In this

framework, changes in the pension system such as changes in the coverage rate, in the accrual of

retirement wealth attributable to continued work, more than the level of retirement wealth at a

given point in time, are found to influence the average retirement age (Samwick, 1998).

According to the simulations of Gruber and Wise (2002), a reform that delays benefit eligibility

by three years would likely reduce the proportion of men aged between 56 to 65 out of the labour

force between 23 to 36%.

Mitchell and Fields (1983) apply an ordered logit model to estimate the impact on the average

retirement age of changes in the expected income. Not surprisingly they find a negative impact of

4The so-called "added worker" effect implies an increase in the participation rates when the expected income of thefamily deteriorates (Pissarides 2000).

7


8/44

a rise in social security streams on the average retirement age. The impact of a 10% increase in

the social security benefits was estimated to reduce a retirement age by -0.07 years for all

individuals without any restriction on age. In case of individuals at the age of 60 the effect is

more pronounced when reducing the average retirement age by -0.19 years.

Bottazzi, Jappelli and Padula (2006) estimate - separately for males and females - the impact of

the Italian pension reform on the expected retirement age, omitting the transitional 1993-1997

period of the reform. While their regressions indicate that the patterns found for women are the

same as for men, still the effect on women is somewhat larger. The estimated impact on the

expected retirement age is about 0.7 years for both male and female private sector worker. In case

of public employee and self employed the effect is even higher reaching values over 1 and 2 years

respectively.

Some EU countries have switched from defined benefit to defined contribution pension systems

or at least introduced one pension pillar based on this assumption. Such change may lead people

to stay longer in the labour market and, therefore, is expected to increase the average retirement

age. Friedberg and Webb (2005) support this hypothesis by estimating that employees with

defined contribution plans usually retire one or two years later compared to employees with

defined benefit plan. Furthermore, Diamond (2005) argues in favour of pension systems with low

implicit tax on continued work after the age at which retirement benefits can first be claimed.

Usually low implicit taxes are ensured with a defined contribution system.

Palmer (1999) proposes a notional defined contribution pay-as-you-go system. As usual in

prevailing pay-as-you-go systems, working people contribute to the system providing resources

for contemporary pensioners. However, differently from the DB system, the more people

contribute to system the higher is their future pension. Finally, the rate of return is not affected

by the developments of the financial markets, but by the overall performance of the economy.

So, the system should stimulate people to postpone their exit from the labour market and, in

passing, to its financial stability.

Bloom, Canning and Moore (2007) show that the optimal response to dealing with the solvency

problems that arise in social security when life expectancy increases is to reduce contributions

and increase benefit rates, maintaining solvency exclusively by increasing the retirement age. This

response can maintain solvency because raising wages over time and compound interest on

accumulated savings mean that longer working lives tend to create more than proportional wealth

at retirement.

8


9/44

The retirement age has stabilised and recently partially reversed its declining trend. Again, several

factors have to be taken into account. First, under the pressure of ageing and the medium- to

long-term risks for the financial sustainability of social security systems, several member states

have enacted reforms of the pension systems that have tightened the eligibility conditions for

pension benefits (e.g. minimum years of contributions, retirement age) and reduced their

generosity. Second, some reforms have shifted part of the financial risks from state to employers

and employees. Thus, longer life expectancy and less generous pension benefits may have

induced workers to work longer to accumulate precautionary savings for their old age (i.e. they

have made the income effect prevail over the substitution effect). The next section reviews more

in depth the pension reforms enacted in the member states in the last decade.

5.

Overview of early retirement and pension reforms undertaken in the EU over the1997-2007 period5

Reaching low levels of inactivity among older workers and promoting longer working lives are

key factors to alleviate the negative impact of population ageing on employment and economic

growth (European Commission-EPC 2009 Aging report). The 2001 Stockholm European

Council stressed the importance of reforms encouraging higher employment and participation

rates, especially among women and the elderly; it emphasised that pension reforms are needed to

ensure both the long-term financial sustainability and a certain degree of intergenerational

fairness.

In response to pressures stemming from ageing populations and persisting low participation

rates, all countries of the EU have reformed their pension systems. These reforms comprise a

number of different measures (Table 6 andTable 7) that were meant to keep the sustainability of

public finances mainly by transferring part of the demographic risk from the state to individuals

and by giving strong incentives for working longer.

A widely accepted distinction is between parametric and systemic reforms. Parametric are those

reforms which involve adjustments to the parameters of defined benefit (DB) and pay-as-you-go

5 This section briefly describes the main elements of the reform strategies adopted in the EU27 over the period1997-2007. Information on pension reforms adopted in the EU27 in the years 2000 to 2007 is taken from theLABREF database (http://ec.europa.eu/economy_finance/db_indicators/db_indicators8638_en.htm). For reformsenacted during the Nineties in the EU15, we used the Fondazione Rodolfo De Benedetti database, available at:http://www.frdb.org. Concerning Bulgaria and Romania, for the time being LABREF only covers the years 2003 to2007. Missing information was mainly obtained from Disney, R. (2003), "Public Pension Reforms in Europe:Policies, Prospects and Evaluation", a number of ILO and ISSA papers, as well as the Joint Reports on Social

Protection and Social Inclusion, 2007 and 2008 editions, and the Synthesis report on adequate and sustainablepensions 2006, all available at: http://ec.europa.eu/employment_social/spsi/index_en.htm.

9
http://ec.europa.eu/economy_finance/db_indicators/db_indicators8638_en.htmhttp://www.frdb.org/http://ec.europa.eu/employment_social/spsi/index_en.htmhttp://ec.europa.eu/employment_social/spsi/index_en.htmhttp://www.frdb.org/http://ec.europa.eu/economy_finance/db_indicators/db_indicators8638_en.htm


10/44

(PAYG) public pension systems, without changing their financing mechanisms. Systemic reforms

move away from the PAYG DB- system and adopt a DC-type personalised accounts system -

thus linking more strictly pension contributions to pension benefits.6

The majority of pension reforms adopted in last ten years were parametric, mainly strengthening

the links between contributions and benefits (notably by extending the period over which

earnings are taken into account for benefits calculation) and stricter conditions foreligibility to

first pillar defined-benefit pension schemes (notably through higher retirement ages). For

example, the reference contribution period and wages used for the calculation of old-age

pensions were extended in Finland in 2003; the annual pension accrual rates were also modified

to discourage early exits from the labour market and to financially reward long working careers; it

was also decided that starting from 2009 pensions would begin to reflect changes in average lifeexpectancy. 7 In Finland and Sweden, greater flexibility was given to older workers to decide their

retirement age (abolition of the general retirement age at 65). In Austria, the 2003 pension reform

raised the retirement age to 65 for men (60 for women) starting from 2017, extended the

assessment period for pension calculation gradually from 15 to 40 years and gradually reduced the

accrual rate.8 Finally, the reform of the public old age pension scheme introduced in Portugal in

2000 increased to 40 yearsthe contribution period for a full pension for the private sector9. Other

measures included changes in the taxation of contributions and benefits, or in the pension

coverage, as well as the setting-up and development of mandatory and/or voluntary second- and

third-tier pension schemes.

Almost all countries increased the statutory retirement age, the majority opting for a smooth

transition towards higher retirement ages ( Table 8 ). The age of eligibility to a state pension was

6 The distinction between parametric and systemic reforms is largely used by the international academic community,notably the IMF and the OECD (see for instance "Pensions at glance", OECD, June 2007). The key parameters of

DB pension schemes can be grouped into: income measures (ceiling or other restrictions on pensionable earnings;number of past salaries included in the calculation of the pension; revalorization mechanism for past salaries);eligibility conditions (statutory retirement age, minimum retirement age (for early retirement), minimum vestingperiod, contribution rate); benefit formula; (accrual rate; reduction factors for retirement prior or after thestatutory retirement age; maximum replacement rates and/or pensions; minimum replacement rates and/orpensions; indexation mechanism for pensions). The main difference between DB and DC pension schemes lies inthe sharing of risks for longevity between the current generation and future ones - i.e. the shift to DC structure insystematic reforms implies greater risks for individuals.7Germany, Finland and France introduced part-time work before the standard retirement age. In Sweden,individuals can continue working, taking a part-time pension and accrue additional unlimited pension rights. Gradualretirement was introduced in Luxembourg for the employees agreeing to switch from full-time to part-time work.8 One year later, the 2004 reform redesigned the calculation of pension benefits leading to a much stronger linkbetween contributions and benefits, including a bonus/malus system for deferred/early retirement, and introduced a

uniform pension law for all professions.

10


11/44

progressively increased from 65 to 67 in Denmark, Sweden and Germany, in the latter with a very

long phasing-in period. In the UK, the earliest age to take a pension was raised from 50 to 55 in

2004 and a default retirement age was fixed at 65 in 2005, with unjustified retirement ages below

65 years being prohibited. The retirement age was also progressively increased in the Czech

Republic (2003) up to 63 years for men and childless women (women get one-year bonus per

child varying between 59 and 62 years), in Hungary (1997) up to 62, Slovenia (1999) and Romania

(2000). In Cyprus, the retirement age for civil servants was increased from 60 to 63, the same as

in the private sector (where retirement ages range between 63 and 65). In Portugal it was raised

from 60 to 65. The age at which women can receive a first pillar pension was equalised with

men's age in most countries.

Pension reforms involved a systemic change in the financing of the insurance system in few casesonly, notably leading to the conversion of pre-existing DB first pillars into notional defined

contribution (NDC) public pension schemes (e.g. PL, SE),10 or to the introduction of statutory

funded pension schemes (e.g. HU, EE, LV, SK). Some countries (HU, SE, PL, LV, EE, LT and

SK) switched part of the public defined-benefit pension system into funded defined-contribution

schemes, where the pension depends on contributions and interest earned on them.

Systemic reforms were also introduced in countries that established state-supported second and

third-pillar voluntary funded pension schemes, supplementing a gradual reduction of first-pillar

pension levels (Germany in 2000) or promoted third pillar pension funds based on employees'

own savings (France in 2003). Several countries encouraged supplementary pension schemes

either through tax incentives or adjusting contribution rates in the direction of private and

occupational schemes (e.g. HU, DE, NL) so as to promote the development of privately-

managed, fully-funded occupational pensions. Similarly, the automatic transfer of the end-of-

service allowance to occupational pension funds was decided in Italy in 2004.

The changes introduced In several countries were rather incremental building upon previous

reforms dating in some cases from the early Nineties (e.g. Italy). Reforms generally involved the

establishment of stronger actuarial links between benefits and contributions - mainly through

9In 2005, it was extended to employees in the public sector. The benefit formula was again significantly changed in2007.10In Poland, pre-existing defined-benefit PAYG pension scheme was replaced in 1999 by a three pillar systemincluding a notional defined-contribution (NDC) first pillar linking contributions to future pensions, a second pillarthat capitalises individual contributions and is mandatory for the younger generations, and a voluntary third pillarbased on company plans or other savings vehicles.10 Following the shift of the public pension pillar from defined

benefit to notional defined-contribution accounts, the pension benefits depend on contributions made, but thenotional interest rate is set by government and the schemes remain pay-as-you go financed. Similar reforms werepassed also in Sweden (1999), Latvia (1996) and Italy (1993, with very long implementation schedule).

11


12/44

longer contribution periods required for a full pension - and increased incentives for workers to

retire later, notably by means of actuarial reductions for early pensions and increases in pension

rights for deferred retirement.

With few exceptions (e.g. Slovakia), the major reforms in the new Member States were legislated

in the 1990s (for instance, Poland, Estonia, Latvia, Lithuania and Slovenia). In some EU10

countries, recent reforms have increased the generosity of the system, for instance by introducing

new early retirement schemes where they did not exist any more (e.g. in Lithuania, where the

early retirement scheme was abolished in 1995 and re-introduced in 2004 for the long-term

unemployed, the Czech Republic, where a new early-retirement programme in the steel industry

was introduced in 2000) or by reinforcing them (e.g. in Hungary), to help absorb the shocks of

ongoing employment restructuring and economic change.

To take better account of future demographic changes, a significant number of countries

introduced a demographic adjustment in their first pillar pension formula linking pensions to

changes in average life expectancy. This is a common feature of all countries having introduced

systemic reforms, where pensions will in future automatically adjust to changes in life expectancy,

but similar adjustment mechanisms have also been built into systems which have not undergone

systemic reforms (e.g. with the reforms of 2003 and 2004 in France and Germany. Similar

provisions have been introduced in DK, FR, AT, FI, LV, LT and, more recently, in PT (2007).

Discouraging early retirement...

Early retirement benefits, which vary by country and usually by professional group depending on

the nature of work, is the main reason for early exits from the labour market. They are often used

as an instrument of employment policy, to artificially lower the unemployment rate of the elderly.

Reducing the generosity of early retirement pensions was a key component of all pension reform.

To discourage early exits from the labour force, Member States have abolished early retirement

schemes, substantially reduced their generosity and introduced bonuses in case of postponement

of retirement for those extending their working lives (Table 8).

For example as part of the 1999 pension reform, in Poland the "pre-retirement allowance" was

discontinued in 2001, while the eligibility conditions for obtaining "pre-retirement benefits" were

made more stringent in 2004.11 A comprehensive reform of the pre-retirement pension system

11 Both schemes had been introduced in 1994 to accompany employment restructuring in the waning branches andoutdated sectors of national economy.

12


13/44

was approved in France in 200312. In Finland (2003-2004), the qualifying age for early old age

pension was raised to 62 and the individual early retirement, available to people with reduced

working capacity aged 60 to 64, was phased-out. The early retirement pension for older long-term

unemployed will be abolished in 200913. Some early retirement schemes were suspended and

abrogated in Portugal in 200514 and the conditions for accessing early retirement tightened in

Czech Republic and Spain (2006). Germany, Hungary, Slovakia (2006) and Portugal (2007) cut

early retirement benefits, raised the minimum contributory period to be eligible for an old-age

pension and tightened the access to schemes open to unemployed. In Latvia, the possibility to

early retire was abolished in 2008. The early retirement age was gradually raised in Austria in

2003, and the possibilities for early retirement will be phased out by 2017. In Germany (2004),

the minimum entry age for early retirement on account of unemployment was increased from 60

to 63. The earliest age at which a private or occupational pension can be taken was also raised in

those countries where this has an impact on the effective labour market exit age (e.g. UK, IRL).

In Sweden (2000), early retired people were allowed to return to work while the tax advantages

for early retirement were abolished in the Netherlands.

Working beyond the official retirement age was supported in many countries for instance with

higher accrual factors e.g. CZ, EE, LU, DE, EL, HU, PT, SI - or with the introduction of

supplements for deferred public old-age pension (e.g. DK). Partial retirement was introduced in

Germany (2001) and the UK (2004) and gradual retirement in France (2006). In this country, a

new form of fixed-term contract for job seekers aged 57 or more was introduced in 2006, while

the so-called 'Deladande Contribution' - a tax to be paid by companies dismissing employees

aged 50 years and over - was gradually phased-out to improve the employability of older

workers15. Incentive schemes for workers who decide to remain in the labour market after the

official retirement age were decided in Italy, France, Spain and the UK.

12 The 2003 reform, which was embedded in the pension package known as the 'Raffarin Act', included limiting fiscalincentives for pre-retirement schemes to physically demanding jobs and restructuring firms in financial distress;eliminating progressive early retirement; increasing the cost of company early-retirement schemes, placingrestrictions on state-financed early retirement. Even so, employers may still require employees who have the right toa full pension to retire between the ages of 60 and 65 if the worker is covered by an early retirement scheme put inplace before the reform came into force or if an extended sector-level collective agreement, providing forcompensatory measures for such retirement, was reached before 1 January 2008. A number of sectors have takenadvantage of this option for maintaining retirement before the age of 65.13 If people become unemployed at the age of 57, they will be entitled to the income-related daily unemploymentallowance until the age of 65 if they have worked for five years during the previous 15. Those born before 1950 willbe entitled to a daily unemployment allowance from the age of 55 until the age of 60; thereafter, early retirement andthen full retirement will be still possible.14 Previously, workers in Portugal could qualify for early retirement benefits either at age 55 with 30 years of

contributions or at age 58 if they were unemployed.15 The Deladande Contribution was introduced in 1987 to compensate for the removal of the administrativeauthorisation of redundancy but in practice obstructed the recruitment of people aged 50 years and older and

13


14/44

6. An empirical evaluation of the effect of pension reforms on the older workers'

participation rates in the short-term

The OECD has conducted an extensive research on the impact of policies and institutions on

employment and unemployment in the OECD countries.16 This work showed that high implicit

taxes on continued work deter older workers from remaining in the labour market, while high

statutory retirement ages have the opposite effect.17 The characteristics of the old age-age public

pension systems (e.g. standard retirement age, accrual rates) and other forms of income support

(early retirement schemes) are found as the main determinants of the differences in the 55-64

participation rates across countries and over time (Blondall and Scarpetta, 1998; Duval, 2003).

In this section we verify the impact of pension reforms on the participation rates of specific

groups of older workers with a difference-in-difference approach. This approach requires the

identification of a specific policy intervention against which one should compare the difference in

outcomes before and after intervention for a treatment and a control group. A source of spatial

and temporal policy variation in the reforms carried out is necessary to estimate this effect.

We exploit the information available from LABREF and other sources (e.g. FRDB, MISSOC etc)

to identify a chronology of reforms.18 Reforms are classified in three categories. First,

fundamental reforms are those systemic reforms that imply a change from defined benefits to

notional defined contribution first pillar pension schemes or that transfer public pension savings

partly to private funded schemes. To this category belong parametric reforms that entail a change

in the eligibility conditions (e.g. statutory retirement age, years of contributions). These reforms

are usually gradually phased in and imply long implementation lags. Second, measures that do not

modify financing or eligibility conditions are deemed as non fundamental, namely those

transferred possible redundancies to employees who were soon to reach 50 years of age. The contribution will be

phased out completely in 2010.16 Bassanini, A. and R. Duval (2006), "Employment Patterns in OECD Countries: Reassessing the Role of Policiesand Institutions", OECD Economics Department Working Papers, No. 486, OECD Publishing17 A 10 percentage points cut in the implicit tax and a one-year increase in the standard retirement age are estimatedto raise the employment rate of older workers by 1 and 0.6 percentage points, respectively.18LABREFprovides information on reforms enacted in various years by the 27 Member States. It is an inventory of

labour market reforms jointly managed by DG ECFIN and the Economic Policy Committee. It is conceived as atool to provide comprehensive description of qualitative features of the reform process, including the design ofenacted reforms, their scope and durability. To date, the database covers the years 2000-2006 for the EU27.Information for the year 2007 will be made available to the public in April 2008. The database can be freelyaccessed at: http://europa.eu.int/comm/economy_finance/indicators/labref_en.htm. For a description ofLABREF see European Economy Research Letter Vol. 1, issue 3 November 2007. As regards pension reformsLABREF provides information distinguishing policy measures in the area of Disability benefits, Early retirement

schemes, Contributions , Coverage, Eligibility conditions, Level and tax treatment of pension reforms. For theyears 2000-2006, the chronology of pension reforms is taken from LABREF. For the previous years theinformation draws on different sources (e.g. EIRO, MISSOC, NATLEX).

14
http://europa.eu.int/comm/economy_finance/indicators/labref_en.htmhttp://europa.eu.int/comm/economy_finance/indicators/labref_en.htm


15/44

modifying the tax regime of contributions and pension benefits, indexation rules, or introducing

second and/or third pension pillar gradually and on a voluntary basis. The third group gathers all

measures implying phasing-out of early retirement schemes.

Graph 3 displays the cumulated number of fundamental, non-fundamental pension and early

retirement reforms for the period 1990-2006. Three things emerge. First, an increasing number of

countries introduced reforms that changed the philosophy of the system (fundamental reforms).

As of 2006 nearly every European country, especially of the EMU (Table 6), had reformed its

pension system. Second, starting from 2000, non-fundamental reforms are more frequent than

fundamental or early retirement reforms. Third, early retirement reforms rare in the 1990s

became more frequent in the early 2000s.

This rich variation in policies across countries and over time can be exploited to assess their

effect on the older workers participation rates. Each measure is considered a discrete event

which occurred at a specific point time for each country. The value of a variable of interest after

certain legislation has taken place is compared to its value before such a change occurred. To

control for factors unrelated to specific policy intervention, the before-after comparison is

evaluated against the average of a control group.

In the period under consideration almost all countries undertook a pension reform. The quasi-

natural experiment framework requires that pension reforms are a source of exogenous variation

with respect to shocks to the participation rates. Consistently with the common belief (Peerson

and Svenson), we assume that the main motivation for governments to undertake a pension

reform is to achieve financial sustainability of social security rather than to offset trends in

participation rates and in the retirement age.

Our sample covers 27 countries over the period 1990-2006. 19 To define our treatment group we

identify as reform year the year in which a reform is enacted. When reforms of the same type arepassed in two consecutive years we treat them as a single event; the average participation rate is

taken as representative of the participation rate at the time of the reform. Similarly, if there are at

most two years between two years of reforms we treat them also as one event. Our control group

is made out of the remaining periods. Within both groups we compute the average change in the

participation rate. Finally, the average change in the participation rate of the treatment group is

19 Since data on participation rates from European LFS Statistics are not available for all years for all countries thepanel is unbalanced.

15


16/44

compared with average participation rate for the control group. If a reform is successful, the

difference between the participation rates of the two groups should differ from zero.

One way to detect this is to compare the change in the participation rate 1, 2 and 3 years after a

pension reform has been implemented with the change in the participation rate in all periods but

those that followed a reform. The change in the participation is modelled as follows:

tititi vIPR ,,, += ; Iit equals 1 if country i enacts a reform in period t and zero otherwise. A

similar expression holds for a country j with ji. The average change of the participation rate in

reforming years relative to change of the participation rate in years of non reform can be written

as follows =

IT

T

t

I

i

JS

PRPRS

s

J

j

sjti ,,

. The reform in countryi is successful if is

statistically different from zero. We evaluate the effect of pension reforms comparing the average

change in the participation rate after a pension reform with the average change of the

participation rate over the sample period excluding those years where a reform occurred.20

For each target group, the first two columns ofTable 1 toTable 3, report the average change in

the participation rate over reforms and non-reforms years; the statistical significance of their

difference appears in column 321. Table 1 suggests that compared to the non-reform years the

participation rate of the 50-54 and 60-64 age groups rise significantly in the years near to the

reform year. Conversely, no significant change is detected for the participation of those belonging

to the 55-59 age group. While fundamental reforms do not have significant effect on the

participation rates in the years just following the enactment of the reform, probably because of

the gradual phasing-in (table 2), parametric reforms entail a change in the participation rate of

those with age between 55 and 59.

20 In contrast, we do not look at the effect on the participation rate of changes in one specific element of the system(i.e. contributions, eligibility conditions, retirement age, indexation formula, and the like). We leave this for future

work.21 Since it may take some time for a pension reform to have visible effects on the participation rate, we calculated theaverage change in the participation rate over a period of 6 years following a pension reform.

16


17/44

Table 1 - Average annual change of the participation rate after EARLY RETIREMENT reforms' years and yearswhere no reforms occur

No reforms years Reforms years z-test: same mean

changesParticipation rate 50-54 0.5 0.9 1.9Participation rate 55-59 0.7 0.9 0.6Participation rate 60-64 0.3 0.9 2.4Source: Authors calculations on LABREF database; the difference between the participation rates of the no-reforms and reforms years is statistically different from zero at 5% of confidence when the value of the z-testis above 2

Source:Commission services.

Table 2 - Average annual change of the participation rate after FUNDAMENTAL reforms' years and years where

no reforms occur

No reforms years Reforms years z-test: same meanchanges

Participation rate 50-54 0.8 0.5 -1.4Participation rate 55-59 0.7 0.9 0.4Participation rate 60-64 0.3 0.6 1.1Source: Authors calculations on LABREF database; the difference between the participation rates of the no-reforms and reforms years is statistically different from zero at 5% of confidence when the value of the z-testis above 2


Table 3 - Average annual change of the participation rate after NON-FUNDAMENTAL reforms' years and yearswhere no reforms occur

No reforms years Reforms years z-test: same meanchanges

Participation rate 50-54 0.6 0.5 -0.4Participation rate 55-59 0.4 1.1 2.1Participation rate 60-64 0.2 0.5 1.2Source: Authors calculations on LABREF database; the difference between the participation rates of the no-reforms and reforms years is statistically different from zero at 5% of confidence when the value of the z-testis above 2


Graph 4 shows the time pattern of the participation rate around the reform event for the three

reforms types and the three age groups. We consider only those reforms that are followed at

least by one year; hence, measures taken in 2006 are excluded from the sample. Next, in order to

select the reform years we treat two consecutive periods of reform as a one reform year. The

same rule applies for years once there are at most two years between two years of reforms.

Consequently, the participation rate in the selected years is calculated as a simple average in these

years.

17


18/44

The figure plots the average change in the participation rate compared with the year in which the

reform occurred. Hence, each point represents the cumulated change up to and since the

enactment of the reform. A successful reform implies a change in the slope in the years that

follow. Before the pension reform, all groups have participation rates lower than or as big as the

rate observed in the year when it is enacted. Then the participation rate increases, and after 3

years it is on average 5 percentage points higher than at the year of enactment.

Graph 4 shows the cumulated change of the participation rates before and after the enactment of

early retirement, fundamental and non-fundamental reforms.22 The following points are

noticeable:

The increase in the participation rate is mainly due to the female component, with

increases dominated by a long-term trend.

After early retirement reforms, the participation rate of women aged 55-59 slightly

accelerates, while the profile of the men rate is more muted.

The change in the participation rates of the oldest group barely differ by gender.

The 50-59 male participation rate changes after early retirement reforms.

Non-fundamental reforms modifies the 55-59 participation rate

The profile of participation rates does not change when fundamental reforms are enacted,

which is consistent with these reforms being usually gradually phased in.

The profile of female participation rate does not change in response to any type of

reform. Yet, we dont consider this an evidence of their ineffectiveness as female

participation is dominated by a long-term trend unrelated to reforms of social security.

These findings are suggestive of a positive impact of early retirement reforms on the participation

rate of specific groups of older workers. The different response for the male and female rates is

consistent with differences in the elasticity of the labour supply to the implicit tax rates and in the

length of working careers and years of contribution to social security. Thus, tightening the access

to early retirement would induce women to postpone retirement.

22 We consider only those reforms that are followed at least by one year; hence, measures taken in 2006 are excludedfrom the sample. In addition, when one reform is followed within four years by another reform of the same type, weconsider in the calculation only the three years preceding and following the first reform.

18


19/44

Of course, participation rates also change in response to the business cycle. In line with the

cyclical ups and downs, those out of the labour force may be induced to starts searching actively

for a job when they perceive that their employment chances have improved. Similarly,

unemployed people may stop searching for a job when their employment prospects weaken and

leave the labour force (the so-called discouraged worker effect). Thus, controlling for the state of

the economy is necessary to identify the effects of pension reforms on the participation rate.

Finally, the participation behaviour is influenced by changes in the socio-economic aptitudes

towards work of the elderly, not necessarily related to governments' interventions. The fact that

participation rates can be influenced by other factors invites shifting to multivariate analysis.

Before proceeding with the analysis an important caveat is needed. Short-term changes in the

participation following a pension reform, as the one considered in this paper, tell nothing aboutthe lags needed for a reform to fully influence the retirement decision and the participation rate.

Pension reforms, especially fundamental, are gradually phased-in and their impact may become

visible only after some years, when an increasing number of cohorts born over successive years

start to be under the new regime. Therefore, the expected gains of pension reforms cannot

always be perceived immediately and their short-run effect is uncertain. Moreover, due to the

gradual phase in, it is unlikely that the oldest generations would change their retirement

behaviour because of the reform. In contrast, those aged between 50 and 54 are more likely to

revise their inter-temporal consumption/leisure allocation. In general, when a reform is

announced, agents may respond with imperfect foresight when two dimensions of uncertainty,

namely the timing and the measures adopted to reform the system, prevail (Butler 1999). Finally,

early retirement and non-fundamental reforms may have shorter implementation lags, and their

effects can be more visible in the short-term. However, delay between announcement and

enactment creates in general the possibility for agents to reassess how the reform will affect their

incentive to retire prior to the effective implementation of the new regime (Santoro, 2006). 23

Thus, the effects of the reforms in the short-term are highly uncertain and depend on how

different cohorts react to current or perspective changes in the rules of the social security system.

For example, for those relatively far from the statutory retirement age, any change in their

participation rate due to the reform would be induced by an announcement effect.

23 Santoro finds unintended announcements effect of the Italian pension reform of September 1992. Santoro, M., M.,(2006), Early announcements of a public pension reform in Italy CBO WP-1

19


20/44

Econometric Methodology

To capture the effect of reforms we estimate a reduced form regression for the participation rate

itiitiitiit

ititit

iitittiit

SEXERSEXNONFUNSEXFUN

ERNONFUNFUN

SEXtrendtrendPR

+++

+++++++=

222

111

21

where PRit is the participation rate for different age groups in country i at time t; i and t are

fixed effects for countries and years respectively, SEXi is a dummy equal to 1 for women and 0

for men; ERit, FUNit and NONFUNit are dummy variables taking the value 1 if a reform occurs

in country s at time t and zero otherwise. , and is the mean difference between countries

that undertook a reform of one of the three types and those that didn't. In practice we compare

the participation rate in countries enacting a pension reform in a given year with the participation

rate in countries that did not enact a pension reform controlling for other (non-reforms factors)

that may influence participation. The unemployment rate ust captures the cyclical components of

unemployment while long-term changes are represented by country/gender specific trends. 24

The reference group in the equation is men. Thus i represents the average (over time) activity

rate of male in country i. Since a reform may imply different effects on the implicit tax rate and

pension wealth of groups with different working histories, we expect a response that differs

across age groups and gender. The interaction between SEX and the reforms dummies would

capture this differential effect. Including interaction of this sort is also convenient when treated

and control group are very similar and/or the treatment and the control group differ along other

dimension of the data, in our case sex; it may also remove trends along these dimensions (Meyer,

1995). To account for lagged effect of enacted reforms we introduced the reform dummies up to

3 lags (i.e. 3 years).

The use of fixed effects allows controlling for unobserved heterogeneity possibly correlated with

the policy dummies. This happens when the participation rates and the decision to undertake

reforms of any type are correlated. Under these circumstances the fixed effect estimator is

consistent and unbiased. In addition to a country specific unobserved component, there can be a

20


21/44

common latent factor which influences both the participation rate and the reform dummy. This

happens when exogenous trends in participation rates (e.g. increase in level of education or

female participation) make a reform of the pension system more likely (for example, because

there is stronger support for reforming the pension system when the participation rate is low

rather than high). In this case the fixed effect estimator is inconsistent and inefficient (Coakley,

Fuertes and Smith, 2004). Conversely, the two-way fixed effects provide consistent and efficient

estimates. In our case, the inclusion of period dummies would absorb all the values of the

coefficients of the reform dummy making them not significant. To avoid this we account for

unobserved common factors with a time trend, which is equivalent to controlling for period

effects when the coefficient of the trend variable is the same across countries.

The introduction of lagged of the reform dummies control for possible correlation between theseand the country specific effects. Finally, to control for the presence of common shocks hitting

men and women in each country we correct standard errors using a robust covariance estimator

according to the formula developed by Liang and Feger (1986)25across groups. We estimate the

equation controlling for fixed effects and for fixed and time dummies26.

Results

Before commenting the results, a note of caution is needed for the relatively limited number of

observations and reforms events. Moreover, it is worth reminding that our analysis focuses only

on the short-term impact of pension reforms, while in many countries these reforms are phased

in only gradually.

The results highlight a different response of the participation rate across gender, age and country

groupings (table 6). Columns 1 and 2 show, respectively for the EU27 and the EMU, the

24We tried specifications with different combinations of common and country specific trends. In light of the stronginstitutional characteristics of European labour markets we preferred to include country and gender/ specific trendsResults are available from the authors.25 This is implemented in Stata with the cluster command. The clustering adjusts for correlations between the errorterms over subgroups. In practice there are less independent observations standard errors should go up. If the errorterms are not independent in a subgroup of observations (such as for the different time periods for a specificindividual in a panel, or e.g. for observations that are spatially close) clustering avoids that common group errorsgenerate too low standard errors (Moulton, 1990)

26 Controlling for period fixed effects would imply that the estimated coefficients would capture all the effects of ourreform dummies which are slowly time varying. Preliminary evidence based on ANOVA F-test suggests that forearly retirement and non fundamental reforms there is more similarity in the number of reforms across time averagesthan across countries averages. The opposite occurs for the number of fundamental reforms with an average which

is more similar across time than countries. This implies that the former types of reforms are enacted in a specificcluster of countries uniformly over time. Conversely the latter are enacted in specific years in a large set of memberstates

21


22/44

estimates of the effects of pension reforms pooling data over the age dimension. The results for

the full sample show an increase in the EU27 participation rate following a fundamental reform,

though the coefficients are not statistically significant. Similar results are obtained for the male

rate when estimates are limited to EMU countries; conversely, the effect on women is negatively

signed, though statistically insignificant. In the case of non-fundamental reforms we have similar

results for the EU27, i.e. positive but insignificant. In contrast, the estimates for the EMU

countries suggest that non-fundamental reforms increase the overall male rate, while the effect on

female participation is ambiguous. Finally, reforms tightening the access to early retirement

increase female participation, more in the EMU than in the non-EMU countries. Conversely,

their effect on male participation is in EMU and non-EMU countries negative or insignificant.

Columns 3 to 7 display the outcome for three age groups. For early retirement reforms, we find aconsistent pattern across different age groups of women. Reforms tightening the generosity of

the early retirement schemes tend to increase the female participation rates, with statistically

significant coefficients, especially for the ages close to the statutory retirement (55-59). 27 By

contrast, the participation rate of men aged 50 to 59 is negatively affected by these reforms. Only

in the case of men belonging to the 60-64 age bracket of the EMU sample, participation increases

after early retirement reforms. In case of reforms that change the main financing characteristics

of the pension system (fundamental reforms), we found a short-term negative impact on female

participation rates in particular for the 55-59 and 60-64 age groups. In the case of men, the

estimates suggest a positive response, in particular for those belonging to the 55-59 bracket.

Those reforms that we have dubbed as non-fundamentals appear to be effective in raising the

participation rate of men both in the EMU and non-EMU, though the coefficients are significant

only for the EMU sample. On the contrary, women participation seems to fall in the short-term.

However, the uncertainty associated to these results is higher probably due to non-fundamental

reforms category being a residual gathering a range of diverse measures. Thus, the implicit

assumption that these different measures have the same impact on the participation rate might

not be valid. Finally, the impact of the early retirement reforms on women is in absolute terms

always the largest. While for men, fundamental reforms seem to have the largest effect on the

participation rate for the central age bracket.

One problem with these estimates is that shocks to the participation rate might also hit the

variable used to capture its cyclical component, i.e. the unemployment rate, implying that the

coefficients measuring its impact on participation are biased downward - as the correlation

27The impact is larger in the EMU sample.

22


23/44

between the shock and the unemployment rate is negative - and inconsistent. To correct for this

endogeneity, the equation has been re-estimated with instrumental variables using the own lagged

values of the unemployment rate as instruments (table 7). As expected, the IV estimates of the

coefficients of the unemployment rate are lower than the OLS estimates. Different specifications

across countries and age groups suggest that the participation rate is broadly more sensitive to the

unemployment in non-EMU countries. One exception is the participation rate of the group 50-

54, which has a response to the difficulty of finding a job due to the cyclical conditions as big as

in the rest of the EU.

Turning to the role of reforms, the IV estimation suggests for both the EU27 and the EMU

sample, a statistically significant and positive effect of fundamental reforms on the overall male

and female participation rates (col 1). When the focus is on specific age groups the effect on themale and female participation rates are respectively positive and negative. For the EU27 sample,

non-fundamental reforms have usually a positive effect on participation rate, which is, however,

precisely estimated only in the case of women. Conversely, for the EMU countries non-

fundamental reforms increase the male participation rate but decrease that of women of age

between 55 and 59. Finally, reforms of early retirement reduce the participation rate of men,

especially those aged between 50 and 54, but increase sizeably that of women.

To account for persistent trends in participation rates unrelated to pension reforms, we includegender and country specific time trends in table 8. Adding trends usually turns out in a lower

impact of reforms, implying that in the specification without trend the impact of reforms is

biased upward, as part of these trends get caught in the reform dummies. Moreover, the fact the

standard error of the coefficients measuring the impact of reforms remained unchanged suggests

that the introduction of specific trends does not introduce multicollinearity that reduces the

precision of the estimates.

Thus, when we control for gender and country specific trends, we find that

fundamental reforms increase the participation rate of older men, respectively in the EU

and the EMU sample, by about 2/3 of and percentage point within 2 years. For both

samples, the response of the male participation rate to fundamental reforms conditioned

to age is hump-shaped: low for the youngest and the oldest of the older workers age

group and high for those with age at about the average retirement age; participation rates

of men aged between 55 and 59 raise by about 2pp. Conversely, female participation

23


24/44

declines in the short-term, more in EMU than in non-EMU countries, offsetting the

overall effect of fundamental reforms.

For the EU sample, non-fundamental increase the overall female participation rate

(+0.7pp in the year of reform), especially of women aged between 50 and 54, while the

male rates remain mainly unaffected. The opposite is found when the estimates are

restricted to the EU subsample. In this case, the male rate increase again the 50-54 age

group being the more reactive while the female components remain mainly unchanged

with the exception of women of age between 55 and 59 whose participation rate drops by

more than 2.5pp.

Early retirement reforms have a positive effect on the female participation rate, especially

for the 55-59 age group of the EMU sample. Conversely, in respectively the EU and the

EMU samples, the male participation rate drops or remains mainly unchanged.

7. Conclusion and policy implication

This paper investigates the short-term effects of pension reforms on the participation rates of

specific age groups belonging to the 50-64 age class with a diff-in-diff approach. Variation across

countries and time in pension reforms enacted in the member states provides the information

needed to examine the effects of these reforms.

The descriptive and preliminary econometric analysis conducted on a sample of 27 EU countries

suggests a different short-term impact of pension reforms on the participation rate of men and

women. Reforms tightening the access to early retirement have a short-term positive effect on

the female participation rate, but reduce somewhat male participation. In our view, these findings

reflect the different length of working life of men and women. A full pension is usually granted

to anyone who has been working for a certain number of years. If someone does not reach the

statutory number of working years, his or her pension is consequently reduced. When men enter

the labour market, they tend to have more stable career path than women and to work

continuously until retirement age (e.g. Hall, 1982). By the official retirement age, males have

worked a sufficient numbers of years to get a full pension. As long as the pension reform reduces

the expected lifetime income, it creates an incentive for those that have accumulated enough

financial wealth to retire earlier. Thus, the optimal retirement age is defined as the upper

threshold such that is never optimal to retire after that age as lifetime income is downward

sloping (B.Jc.Heijdra and Romp, W.E., 2007). Following the announcement of a reform that

24


25/44

makes less generous the pension system, men just below the retirement age may find more

convenient to anticipate the exit decision, not to miss a generous pension. These findings suggest

the risk of a run on pension funds well before the changes take effect. This has been indeed the

case following the announcements of restrictions of early retirement in some EU Member States,

according to the 2009 Commission working document Joint Report on Social Protection and

Social inclusion.

Conversely, women have more career interruptions than men, especially because of maternity

leave and family reasons, and the number of years spent working at the age of retirement is

smaller than men. This difference may explain why the female participation rate raises in

response to early retirement reforms. Compared to men, women have to reach a reasonable

pension or accumulate a sufficient amount of precautionary savings before being able to retirewith (not too large) drop in consumption. The effects are stronger in the EMU than in the non-

EMU countries.

The results for non-fundamental reforms are more uncertain. The positive effect of non-

fundamental reforms for men is not surprising. These reforms usually adjust upwards the

contribution rates, implying a lower net wage. If the substitution effect prevails, an individual

prefers to work more. There is an additional motive for working more, which is related to the

increasing life expectancy. Because of a longer life span an individual needs to work more in

order to accumulate sufficient amount of wealth. As the real wage drops, he/ she needs to work

more to reach an intended level of consumption during the retirement age.

In contrast, reforms that change the way of financing pensions or the eligibility conditions

(fundamental reforms), usually with long phasing-in periods, may have unintended short-run

effect on the female participation rate, especially of EMU countries.

Our findings point at the importance of designing pension reforms and strategies to reform socialsecurity that reduce the risks of undesired effects on the decision to remain into the labour

market. There is plenty of evidence that workers' information about pension rules and

uncertainties about long transition periods may influence in the short-term the retirement

decision in a way which is not consistent with the intended effects of the reform. While transitory

periods may be needed to gain the political support for the reforms, long and reiterated

discussions on how to reform the social security system may add uncertainty and, if allowed by

the rule in force, lead to anticipate the retirement decision even in cases where reforms involve

future and not current older workers. Well-informed individuals are far more responsive to

25


26/44

pension incentives, while ill-informed individuals seem to respond systematically to their

misperceptions of pension incentives (Chan and Huff Stevens, 2008).

To buttress these results, we plan to extend the empirical analysis in five directions. First, in the

regression, we control for the determinants of participation unrelated to reforms with country

fixed effects, period dummies or a common trend. The evidence found needs to be corroborated

by enlarging the set of controls to observable variables, such as self-employed, age of entry into

the labour market, per capita income, share of employee working in the public sector. Second, to

get an indication of the short-term effect of pension reforms on the retirement decision our

result should be validated by similar finding for probability of withdrawing from the labour

market. Third, to better study labour force dynamics in response to pension reforms we need to

combine the cross-country policy variation with individual information on the labour marketstatus. To use individual data from older workers' self-reported satisfaction to investigate the

effect of pension reforms on their retirement decisions. Finally, in the estimate we do not take

into account that for the retirement decision what matters is not the individual income but the

family income. There is evidence for the US of a differential response to policy changes of men

from one earner vs two earner households (Gustman, A. and Steinmeier, T, 2008). Extending the

analysis to the participation rates of married men and women might provide some hindsight on

the different, and sometime puzzling, response of the male and female participation rates to

pension that found in our estimates.

26


27/44

References

- Autor, D. H., Donohue, J.J. III and S.J. Schwab (2004). The Employment Consequences of

Wrongful-Discharge Laws: Large, Small or None at All, American Economic Review 94, 2.

- Bassanini and Duval (2006)

- Bloom D.,Canning D., Mansfield R., Moore M. (2006): "Demographic Change, Social Security,

and Savings", National Bureau of Economic Research, Working Paper, No. 12621

- Bloom D.,Canning D., Moore M. (2007): "A Theory of Retirement", National Bureau of

Economic Research, Working Paper, No. 13630

- Boeri, T. A. Brugiavini and C. Maignan (2001) Early retirement: reasons and consequences. in

T. Boeri, A. Borsch-Supan, A. Brugiavini, R. Disney, A. Kapteyn and F. Peracchi (eds), Pensions,

More Information, Less Ideology: Assessing the Long-term Sustainability of European Pension Systems(Kluwer

Academic)

- Brugiavini A. (2001) "Early retirement in Europe", European Rewiev, Vol. 9, No. 4

- Chan, S. and A. Huff Stevens (2008), "What you don't know can't help you: pension knowledge

and retirement decision-making", The Review of Economics and Statistics, May, 90(2) pp. 253-266.

- Coakley, J., Fuertes, A. Smith, R. (2004), "Unobserved Heterogeneity in Panel Time SeriesModels", Birbeck Working Papers in Economics and Finance 0403.- Colie C., Gruber J. (2000): "Social Security Incentives to Retirement", National Bureau of

Economic Research, Working Paper, No.7651

- Commission Staff Working Document (2008): "Joint Report on Social Protection and Social

Inclusion", Supporting Document

- Diamond P. (2005): "Pension for an Aging Population", National Bureau of Economic

Research, Working Paper, No.11877

- ECB (2007), "Labour supply and employment rates in the euro area countries. Developments

and Challenges", Task Force of the Policy Committee of the European System of Central Banks.

Occasional Paper , No87.

- Friedberg L., (2007) "The Recent Trend Towards Later Retirement", Center for RetirementResearch at Boston College, An Issue in Brief, series 9

- Guber J., Wise D. (2002): "Social Security Programs and Retirement around the World: Micro

Estimation", National Bureau of Economic Research, Working Paper, No.9407

Gustman, A. and Steinmeier, T, (2008). How Does Modelling of Retirement Decisions at the

Family Level Affect Estimates of the Impact of Social Security Policies on Retirement?,Michigan

Retirement Research Center WP No. 179.

Hall, R. E., (1982) "The Importance of Lifetime Jobs in the U.S. Economy", American Economic

Review, 72(3), 716-724.

27


28/44


29/44

Table 4 - Life expectancy at birth

Belgium Bulgaria Czech Denmark Germany Estonia Ireland Greece Spain France Italy

1980 73.3 71.1 70.4 74.71

73.1 70.62

73.33

75.3 75.4 : 75.6

1990 76.2 71.2 71.5 74.9 75.4 69.9 74.8 77.1 77 77 77.2

2006 79.5 72.7 76.8 78.4 79.9 73.1 79.7 79.5 : 81 815

1 1986; 2 1989; 3 1985; 4 1985;

Cyprus Latvia Lith. Luxem. Hungary Malta Netherl. Austria Poland Portugal Romania

1980 : : 70.5 74.75

69.1 70.4 76.57

72.7 : 71.5 69.2

1990 : : 71.5 75.7 69.4 776

77.1 75.8 : 74.1 69.9

2006 80.6 70.9 71.1 79.4 73.5 79.5 80 80.1 75.3 78.9 72.65

1986 ;61994 ;

71985

Slovenia Slovakia Finland Sweden UK

1980 : 70.4 74.58

75.8 :

1990 73.9 71.1 75.1 77.6 :

2006 78.3 74.4 79.6 81 :81985

Source:Eurostat.

Table 5 - Average exit age

1984-1990 1991-1999 2000-2006

BE 58.5 59.6 60.2

DK 65.6 64.6 65.8

DE1

61.5 60.8 62.7

GR 62.7 63.4 63.2

ES 63.2 62.3 63.3

FR 59.6 59.3 59.8

IE 63.9 64.7 66.3

IT 60.7 59.8 61.1

LU 62.3 58.9 60.8

NL 60.3 60.7 63.2

AT2

58.3 61.4

PT 65.1 66.2 64.5

FI2

62.5 62.9

SE2 65.4 65.7

UK 62.3 64.3

CY 67.9

CZ3 59.4 61.2

EE3 65.8 67.6

HU4

58.1 61.1LT

565.2 63.8

LV5

61.4 67.1

MT 60.1

PL3

59.6 58.7

SK5

57.4 59.1

SI4

61.1 62.7

BG 63.5

RO6

61.5 62.5

Source:Commission services.1 1985-1989; 2 1996-1999; 3 1998;41997-1998; 5 1999; 6

1998-1999

29


30/44

30

Table 6 Number of pension reforms by a type of a reform and by acountry group

Fundamental Non Fundamental EarlyEU27 56 87 37EMU 36 55 26Non EMU 20 32 11

Source: LABREF; FRDB Database

Table 7 Pension reforms' characteristics

COUNTRY

MODIFYING THE

PARAMETERS OF

EXISTING DBSCHEMES

INTRODUCINGNDC STATUTORY

SCHEMES

INTRODUCING AFUNDED TIER IN

THE STATUTORY

PENSION SCHEME

REFORMINGEARLY

RETIREMENT

DEVELOPINGPRIVATE

OCCUPATIONAL OR

PERSONAL PENSION

PROVISION

OTHER (E.G. TAXATION,CONTRIBUTIONS, PENSION

COVERAGE,INDIVIDUALISATION OF

PENSION RIGHTS)

BE x xDK x x x

DE x x x x

GR x

ES x x x

FR x x x

IE x

IT x x x x

LU

NL x

AT x x x

PT x x x

FI x x

SE x x x x

UK x x x

BG x xCY x

CZ x x x

EE x

HU x x x x

LT x x

LV x x x

MT x

PL x x x

RO x x

SI x

SK x x x x

Source:LABREF; FRDB Database


31/44


32/44


33/44


34/44


35/44

ROMen: 63 in 1st quarter of 2007Women: 58 in 1st quarter of 2007

Men: 65Woman: 60

2014

1) Old-Age Pensioof standard retiremunder special or dhandicap prior to opersecuted for polcategories, define2) Early Retiremenretirement age to iby minimum 10 ye3) Partial Early Reretirement age to iby maximum 10 ye

SI

Men: 63 in 2009

Women: 61 in 2008(following gradual increase)

No special early pePossibility of exce

of 58 provided thayears of service

SK Old-Age Pension: 62

This level of retirement age will bereached in 2014 for all populationgroups

2014

1stPillar: No age limembership (10 y2nd Pillar: No agethe 1st pillar is rece

Source: MISSOC Comparative Tables on Social Protection in the 27 Member States of the European Union, in the European Economic Area and in Savailable at: http://ec.europa.eu/employment_social/spsi/missoc_tables_en.htm#table2007; LABREF 2000-2007.
http://ec.europa.eu/employment_social/spsi/missoc_tables_en.htm#table2007http://ec.europa.eu/employment_social/spsi/missoc_tables_en.htm#table2007


36/44

Graph 1 Male and Female age profiles for selected countries

Age profile of the male participation rate: Belgium

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

50 52 54 56 58 60 62 64 66 68 70 72 74

1984

2007

Age profile of the female participation rate: Belgium

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

50 52 54 56 58 60 62 64 66 68 70 72 74

1984

2007

Age prof ile of the male participation rate: Germany

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

50 52 54 56 58 60 62 64 66 68 70 72 74

1984

2007

Age profile of the female participation rate: Germany

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

50 52 54 56 58 60 62 64 66 68 70 72 74

1984

2007

Age profile of the female participation rate: Denmark

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

50 52 54 56 58 60 62 64 66 68 70 72 74

1984

2007

Age prof ile of the male participation rate: Denmark

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

50 52 54 56 58 60 62 64 66 68 70 72 74

1984

2007

Source: LFS

36


37/44

Male and Female age profiles in selected countries

Age profile of the male participation rate: France

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

50 52 54 56 58 60 62 64 66 68 70 72 74

1984

2007

Age profile of the female participation rate: France

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

50 52 54 56 58 60 62 64 66 68 70 72 74

19842007

Age profile of the male participation rate: Ireland

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

50 52 54 56 58 60 62 64 66 68 70 72 74

1984

2007

Age profile of the female participation rate: Ireland

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

50 52 54 56 58 60 62 64 66 68 70 72 74

1984

2007

Age profile of the male participation rate: Italy

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

50 52 54 56 58 60 62 64 66 68 70 72 74

1984

2007

Age profile of the female participation rate: Italy

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

50 52 54 56 58 60 62 64 66 68 70 72 74

1984

2007

Source: LFS

Graph 2 Probabilities of exiting in selected countries

37


38/44

Men: Italy

-0.30

-0.25

-0.20

-0.15

-0.10

-0.05

0.00

50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70

avg 1983-1994

avg 2000-2006

Women: Italy

-0.35

-0.30

-0.25

-0.20

-0.15

-0.10

-0.05

0.00

50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70

avg 1983-1994

avg 2000-2006

Men: Belgium

-0.50

-0.45

-0.40

-0.35

-0.30

-0.25

-0.20

-0.15

-0.10

-0.05

0.00

50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70

avg 1983-1994

avg 2000-2006

Women: Belgium

-0.50

-0.45

-0.40

-0.35

-0.30

-0.25

-0.20

-0.15

-0.10

-0.05

0.00

50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70

avg 1983-1994

avg 2000-2006

Men: Spain

-0.60

-0.50

-0.40

-0.30

-0.20

-0.10

0.00

50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70

avg 1983-1994

avg 2000-2006

Women: Spain

-0.50

-0.45

-0.40

-0.35

-0.30

-0.25

-0.20

-0.15

-0.10

-0.05

0.00

50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70

avg 1983-1994

avg 2000-2006

Source: LFS

38

.Impact van pensioenhervormingen op de participatiegraad van oudere werknemers in de EU

Documents

Transcript of .Impact van pensioenhervormingen op de participatiegraad van oudere werknemers in de EU